# Fine-grained classification on Cars (test)

Source: [Fine-grained classification on Cars (test)](<https://www.sota2.com/research/sota/fine-grained-classification-on-cars-test>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-08T07:11:17.077Z

Evaluation of Fine-grained Classification performance on the test split of the Cars dataset, reporting Accuracy and Adversarial Accuracy.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Fine-grained classification |
| Dataset (including any stated split) | Cars (test) |
| Metric | Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 26 of 26 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Accuracy](<https://www.sota2.com/research/sota/fine-grained-classification-on-cars-test.md?metric=Accuracy>): Higher is better; unit: PERCENT.

- [Adversarial Accuracy](<https://www.sota2.com/research/sota/fine-grained-classification-on-cars-test.md?metric=Adversarial%20Accuracy>): Higher is better; unit: PERCENT.

- [Robust Accuracy](<https://www.sota2.com/research/sota/fine-grained-classification-on-cars-test.md?metric=Robust%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| SR-GNN | 96.1 | Not recorded | 68fa6d911e209131cea3ab16 |
| SotA | 95.7 | Not recorded | 68fa6d911e209131cea3ab16 |
| BYOL+CVSA | 87.13 | Not recorded | 68fa6da51e209131cea3b22a |
| BYOL+DiLo | 86.46 | Not recorded | 68fa6da51e209131cea3b22a |
| InsLoc | 86.38 | Not recorded | 68fa6da51e209131cea3b22a |
| BYOL | 86.36 | Not recorded | 68fa6da51e209131cea3b22a |
| MoCo.v2 | 86.08 | Not recorded | 68fa6da51e209131cea3b22a |
| SimCLR | 85.93 | Not recorded | 68fa6da51e209131cea3b22a |
| LooC\* | 85.9 | Not recorded | 68fa6da51e209131cea3b22a |
| Rel-Loc | 85.8 | Not recorded | 68fa6da51e209131cea3b22a |
| Rot-Pred | 85.74 | Not recorded | 68fa6da51e209131cea3b22a |
| MoCo.v2 | 85.71 | Not recorded | 68fa6da51e209131cea3b22a |
| BYOL | 85.49 | Not recorded | 68fa6da51e209131cea3b22a |
| TAPT | 62.5 | Not recorded | 6a264c5690966ade919be7db |
| MTA | 61 | Not recorded | 6a264c5690966ade919be7db |
| R-TPT | 60.9 | Not recorded | 6a264c5690966ade919be7db |
| CLIP | 58.8 | Not recorded | 6a264c5690966ade919be7db |
| MAC | 58.4 | Not recorded | 6a264c5690966ade919be7db |
| TPT | 57.5 | Not recorded | 6a264c5790966ade919be7e0 |
| Ensemble | 56.1 | Not recorded | 6a264c5790966ade919be7e0 |
| SS-TPT | 56 | Not recorded | 6a264c5790966ade919be7e0 |
| CLIP | 55.8 | Not recorded | 6a264c5790966ade919be7e0 |
| R-TPT | 53.5 | Not recorded | 6a264c5790966ade919be7e0 |
| DOC | 52.9 | Not recorded | 6a264c5790966ade919be7e0 |
| TTC | 52.8 | Not recorded | 6a264c5690966ade919be7db |
| TTC | 52.6 | Not recorded | 6a264c5790966ade919be7e0 |

### SR-GNN

Accuracy: 96.1. Result ID: 698faa903a6c314849faee8a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6d911e209131cea3ab16.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d911e209131cea3ab16.md>)

[Implementation](<https://github.com/ArdhenduBehera/SR-GNN>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### SotA

Accuracy: 95.7. Result ID: 698faa903a6c314849faee89.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6d911e209131cea3ab16.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### BYOL+CVSA

Accuracy: 87.13. Result ID: 698fb3da3a6c314849fb3d73.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6da51e209131cea3b22a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6da51e209131cea3b22a.md>)

Recorded evaluation configuration: {"First-stage pre-training":"COCO","Second-stage pre-training":true,"Protocol":"fine-tuned evaluation"}. This record does not assert independent reproduction.

### BYOL+DiLo

Accuracy: 86.46. Result ID: 698fb3da3a6c314849fb3d72.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6da51e209131cea3b22a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa691c1e209131cea204b1.md>)

Recorded evaluation configuration: {"First-stage pre-training":"COCO","Second-stage pre-training":true,"Protocol":"fine-tuned evaluation"}. This record does not assert independent reproduction.

### InsLoc

Accuracy: 86.38. Result ID: 698fb3da3a6c314849fb3d6f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6da51e209131cea3b22a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69ad1e209131cea2346b.md>)

Recorded evaluation configuration: {"First-stage pre-training":"COCO","Second-stage pre-training":true,"Protocol":"fine-tuned evaluation"}. This record does not assert independent reproduction.

### BYOL

Accuracy: 86.36. Result ID: 698fb3da3a6c314849fb3d70.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6da51e209131cea3b22a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa684a1e209131cea1b1e9.md>)

[Implementation](<https://github.com/deepmind/deepmind-research/tree/byol>)

Recorded evaluation configuration: {"First-stage pre-training":"COCO","Second-stage pre-training":false,"Protocol":"fine-tuned evaluation"}. This record does not assert independent reproduction.

### MoCo.v2

Accuracy: 86.08. Result ID: 698fb3da3a6c314849fb3d6c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6da51e209131cea3b22a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa673f1e209131cea15344.md>)

[Implementation](<https://github.com/facebookresearch/moco>)

Recorded evaluation configuration: {"First-stage pre-training":"COCO","Second-stage pre-training":false,"Protocol":"fine-tuned evaluation"}. This record does not assert independent reproduction.

### SimCLR

Accuracy: 85.93. Result ID: 698fb3da3a6c314849fb3d6b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6da51e209131cea3b22a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67d71e209131cea18de5.md>)

[Implementation](<https://github.com/google-research/simclr>)

Recorded evaluation configuration: {"First-stage pre-training":"COCO","Second-stage pre-training":false,"Protocol":"fine-tuned evaluation"}. This record does not assert independent reproduction.

### LooC\*

Accuracy: 85.9. Result ID: 698fb3da3a6c314849fb3d6e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6da51e209131cea3b22a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69961e209131cea22e05.md>)

Recorded evaluation configuration: {"First-stage pre-training":"COCO","Second-stage pre-training":true,"Protocol":"fine-tuned evaluation"}. This record does not assert independent reproduction.

### Rel-Loc

Accuracy: 85.8. Result ID: 698fb3da3a6c314849fb3d69.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6da51e209131cea3b22a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa639f1e209131ce9ffc68.md>)

Recorded evaluation configuration: {"First-stage pre-training":"COCO","Second-stage pre-training":true,"Protocol":"fine-tuned evaluation"}. This record does not assert independent reproduction.

### Rot-Pred

Accuracy: 85.74. Result ID: 698fb3da3a6c314849fb3d6a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6da51e209131cea3b22a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64261e209131cea02e62.md>)

[Implementation](<https://github.com/gidariss/FeatureLearningRotNet>)

Recorded evaluation configuration: {"First-stage pre-training":"COCO","Second-stage pre-training":true,"Protocol":"fine-tuned evaluation"}. This record does not assert independent reproduction.

### MoCo.v2

Accuracy: 85.71. Result ID: 698fb3da3a6c314849fb3d6d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6da51e209131cea3b22a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa673f1e209131cea15344.md>)

[Implementation](<https://github.com/facebookresearch/moco>)

Recorded evaluation configuration: {"First-stage pre-training":"COCO","Second-stage pre-training":true,"Protocol":"fine-tuned evaluation"}. This record does not assert independent reproduction.

### BYOL

Accuracy: 85.49. Result ID: 698fb3da3a6c314849fb3d71.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6da51e209131cea3b22a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa684a1e209131cea1b1e9.md>)

[Implementation](<https://github.com/deepmind/deepmind-research/tree/byol>)

Recorded evaluation configuration: {"First-stage pre-training":"COCO","Second-stage pre-training":true,"Protocol":"fine-tuned evaluation"}. This record does not assert independent reproduction.

### TAPT

Accuracy: 62.5. Result ID: 6a266a01fbc373592c22aaaf.

[Reporting paper](<https://www.sota2.com/research/paper/6a264c5690966ade919be7db.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75cf1e209131cea69f7f.md>)

Recorded evaluation configuration: {"Category":"Tuning-based","Backbone":"CLIP-ViT-B/32","Attack Protocol":"PGD-100 (ϵ = 4.0)"}. This record does not assert independent reproduction.

### MTA

Accuracy: 61. Result ID: 6a266a01fbc373592c22aab1.

[Reporting paper](<https://www.sota2.com/research/paper/6a264c5690966ade919be7db.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73191e209131cea5a896.md>)

[Implementation](<https://github.com/MaxZanella/MTA>)

Recorded evaluation configuration: {"Category":"Tuning-free","Backbone":"CLIP-ViT-B/32","Attack Protocol":"PGD-100 (ϵ = 4.0)"}. This record does not assert independent reproduction.

### R-TPT

Accuracy: 60.9. Result ID: 6a266a01fbc373592c22aab0.

[Reporting paper](<https://www.sota2.com/research/paper/6a264c5690966ade919be7db.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7add1e209131cea8590d.md>)

[Implementation](<https://github.com/TomSheng21/R-TPT>)

Recorded evaluation configuration: {"Category":"Tuning-based","Backbone":"CLIP-ViT-B/32","Attack Protocol":"PGD-100 (ϵ = 4.0)"}. This record does not assert independent reproduction.

### CLIP

Accuracy: 58.8. Result ID: 6a266a01fbc373592c22aaae.

[Reporting paper](<https://www.sota2.com/research/paper/6a264c5690966ade919be7db.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69581e209131cea21ad0.md>)

[Implementation](<https://github.com/OpenAI/CLIP>)

Recorded evaluation configuration: {"Category":"Baseline","Backbone":"CLIP-ViT-B/32","Attack Protocol":"PGD-100 (ϵ = 4.0)"}. This record does not assert independent reproduction.

### MAC

Accuracy: 58.4. Result ID: 6a266a01fbc373592c22aab3.

[Reporting paper](<https://www.sota2.com/research/paper/6a264c5690966ade919be7db.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a264c5690966ade919be7db.md>)

Recorded evaluation configuration: {"Category":"Tuning-free","Backbone":"CLIP-ViT-B/32","Attack Protocol":"PGD-100 (ϵ = 4.0)"}. This record does not assert independent reproduction.

### TPT

Accuracy: 57.5. Result ID: 6a266b10fbc373592c22b2fa.

[Reporting paper](<https://www.sota2.com/research/paper/6a264c5790966ade919be7e0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d691e209131cea399f9.md>)

[Implementation](<https://tinyurl.com/yr3zmhma>)

Recorded evaluation configuration: {"Backbone":"CLIP-ResNet50","Epsilon":"1.0","Augmentation views":"15"}. This record does not assert independent reproduction.

### Ensemble

Accuracy: 56.1. Result ID: 6a266b10fbc373592c22b2fb.

[Reporting paper](<https://www.sota2.com/research/paper/6a264c5790966ade919be7e0.md>)

Recorded evaluation configuration: {"Backbone":"CLIP-ResNet50","Epsilon":"1.0","Augmentation views":"15"}. This record does not assert independent reproduction.

### SS-TPT

Accuracy: 56. Result ID: 6a266b10fbc373592c22b2ff.

[Reporting paper](<https://www.sota2.com/research/paper/6a264c5790966ade919be7e0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a264c5790966ade919be7e0.md>)

Recorded evaluation configuration: {"Backbone":"CLIP-ResNet50","Epsilon":"1.0","Augmentation views":"15"}. This record does not assert independent reproduction.

### CLIP

Accuracy: 55.8. Result ID: 6a266b10fbc373592c22b2f9.

[Reporting paper](<https://www.sota2.com/research/paper/6a264c5790966ade919be7e0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69581e209131cea21ad0.md>)

[Implementation](<https://github.com/OpenAI/CLIP>)

Recorded evaluation configuration: {"Backbone":"CLIP-ResNet50","Epsilon":"1.0"}. This record does not assert independent reproduction.

### R-TPT

Accuracy: 53.5. Result ID: 6a266b10fbc373592c22b2fe.

[Reporting paper](<https://www.sota2.com/research/paper/6a264c5790966ade919be7e0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7add1e209131cea8590d.md>)

[Implementation](<https://github.com/TomSheng21/R-TPT>)

Recorded evaluation configuration: {"Backbone":"CLIP-ResNet50","Epsilon":"1.0","Augmentation views":"15"}. This record does not assert independent reproduction.

### DOC

Accuracy: 52.9. Result ID: 6a266b10fbc373592c22b2fd.

[Reporting paper](<https://www.sota2.com/research/paper/6a264c5790966ade919be7e0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/691565ea0052cac1dc65d709.md>)

Recorded evaluation configuration: {"Backbone":"CLIP-ResNet50","Epsilon":"1.0","Effective batch size":"1"}. This record does not assert independent reproduction.

### TTC

Accuracy: 52.8. Result ID: 6a266a01fbc373592c22aab2.

[Reporting paper](<https://www.sota2.com/research/paper/6a264c5690966ade919be7db.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa774a1e209131cea724cc.md>)

[Implementation](<https://www.sota2.com/here>)

Recorded evaluation configuration: {"Category":"Tuning-free","Backbone":"CLIP-ViT-B/32","Attack Protocol":"PGD-100 (ϵ = 4.0)"}. This record does not assert independent reproduction.

### TTC

Accuracy: 52.6. Result ID: 6a266b10fbc373592c22b2fc.

[Reporting paper](<https://www.sota2.com/research/paper/6a264c5790966ade919be7e0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa774a1e209131cea724cc.md>)

[Implementation](<https://www.sota2.com/here>)

Recorded evaluation configuration: {"Backbone":"CLIP-ResNet50","Epsilon":"1.0","Effective batch size":"1"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
